Analysis of a Landslide on a Railway Track Using Laser Scanning and FEM Numerical Modelling

Abstract

In this study, we present an analysis of the causes of a landslide along a railway track in the Polish Lowlands. The landslide damaged the railway track and caused significant material losses. Digital models of the terrain surface before and after the landslide were elaborated. Remote sensing using LIDAR aerial technique and a terrestrial laser scanner was performed to determine the morphology. Soil mass behaviour was analysed by 3D numerical simulation. A numerical model was created based on geotechnical tests. Taking into account the behaviour of the dry and wet models, the numerical simulation showed the most probable scenario of mass movement. The main reasons for the landslide were rainwater infiltration in the track basement and the unfavourable morphology of the area on which the railway embankment was located. The study demonstrates that combined methods—laser scanning, geotechnical testing of the soil material, and 3-dimensional numerical simulation—enabled the assessment of the causes of the analysed landslide.

1. Introduction

Landslides along transportation routes often occur due to natural and anthropogenic factors [1,2,3,4,5,6,7,8]. In such cases, the influence of human activity can directly or indirectly lead to a landslide in a given location. The anthropogenic causes of landslide formation may be the undercutting of a slope combined with excessive loads at the edge by infrastructural objects. An intermediate influence is often manifested by changes to the surface shape of the surrounding areas–especially concerning the ground surface becoming steeper, or deforestation. Indirectly, a change in land use and re-directing the water flow on slopes are also among the supplementary causes. Furthermore, dynamic impact on the ground in the form of different vibrations is also a common contributor to landslide activation.

In Poland–due to the natural morphology of the ground–there are many landslides (Figure 1), especially in the transport corridors [9,10]. In most, several natural factors contribute to the formation of landslides; these are the geological structure, the terrain morphology, and intensive and long-term rainfalls [8].

Water is the leading natural factor activating landslide movements. Soil becomes saturated with water, which increases its weight as the water displaces the air from the free space between the soil particles; there are numerous reasons for this saturation, but unsurprisingly, rainfall is the primary reason for this.

Experience shows that the most dangerous types of rainfall are not those that are very intensive but, rather, prolonged periods of less intensity [11,12]. This kind of rainfall contributes to increasing moisture, pore pressure and soil plasticity. When the ground is filled with water, its properties change significantly. Changing soil parameters under the influence of water significantly impacts their behaviour in terms of pore pressure. This is of great importance when a landslide is activated. Water infiltration significantly affects the level of groundwater. Frequent fluctuations in the groundwater level (alternating saturation and drying) contribute to a reduction in soil strength, which facilitates ground movement [13].

Figure 1. The location of the research area concerning landslide areas in Poland (red spots) (www.pgi.gov.pl, accessed on 30 March 2022).

Figure 1. The location of the research area concerning landslide areas in Poland (red spots) (www.pgi.gov.pl, accessed on 30 March 2022).

Ground surface monitoring is quite frequently used to reduce the risk of landslides. One monitoring method is the remote sensing technique. A terrestrial laser scanner gives good results for monitoring earth structures, as well as for railway and road embankments [14,15,16,17,18,19,20,21].

In this study, we present an analysis of the causes of a landslide along a railway track in the Polish Lowlands. This type of area is not considered to be prone to the formation of land movements. In the north of Poland, landslides are rare and usually occur on slopes next to rivers. In this case, specific conditions for the activation of mass movements of soil occurred. We performed an analysis of a landslide along a railway track using morphology derived from laser scanning, geotechnical analysis of the constituent soil material, and 3-dimensional finite element analysis. The analysis is common practice nowadays; however, as the study is on a railway track, it has significant social importance.

2. Landslide Area and Methods

The landslide occurred across a railway track near the town of Kwidzyn, which is located in the geologically defined area of the Polish Lowlands. This area is not considered prone to the formation of land movements (Figure 1). Much more dangerous areas from this point of view are in the south of Poland. In this case, specific conditions occurred resulting in the activation of mass movements of soil.

The research area is under the influence of all glaciation in Poland. There were three glaciations in Poland during the Pleistocene. The last of these was the Vistulian glaciation (Leszno phase) [22]. The thickness of the Quaternary deposits is, therefore, quite large in this area (Figure 2).

The Quaternary deposits comprise the glaciation sediments in Poland’s south, central, and central–north areas. They are separated by the Emski and Mazowiecki interglacial sediment settlements and the Holocene interglacial. The thickness of Quaternary deposits varies from 70 m to 250 m. These are mainly sand, gravel sands, individual clay layers, loam, silt and peat.

In June 2020, damage to the railway track occurred due to a landslide. The event resulted in the complete failure of the railway embankment (Figure 3).

It should be mentioned that in June 2020, the monthly rainfall exceeded 200 mm. The rainfall for June 2020 according to the Ringers system is presented in Figure 4. Several meteorological IMGW-PIB stations indicated that the period under consideration was the wettest since 1951. In the Kwidzyn region, rainfall was high at above 100 mm per month (Figure 4). The soil moisture was also high−above 60%. This abnormal amount of rainfall and the high soil moisture caused this hazardous landslide to occur.

Two types of laser scanning measurements were employed: aeroplane LIDAR (Light Detection and Ranging) scanning and terrestrial laser scanning. The first was used to elaborate the morphology model before the landslide occurred. The second was used to analyse the landslide morphology, especially in comparison against the results of the numerical calculations.

LIDAR data from the Airborne Laser Scanning (ALS) represent the terrain in the form of a cloud of measurement points with specific XYZ coordinates. In Poland, the entire country is covered by data from ALS measurements with different point densities, from 4 points/m² to as high as 20 points/m² (in the case of cities). The vertical resolution of the measurements is about 15 mm.

Terrestrial laser scanning was performed in four locations using a RIEGL VZ-400. The measurement points were located at distances no greater than 400 m apart. The vertical resolution of the measurements was about 5 mm. Each measurement produced clouds of points in 3D with an accuracy of approx. 2–5 mm. GPS marked out each measuring point. The point cloud was developed in a RiSCAN PRO [24] because four scans were submitted. The point cloud is oriented in the local coordinate system. In the RiSCAN PRO program, the vegetation was removed within the analysed area.

A 3D numerical model of the landslide was created based on geotechnical studies using the MIDAS FEM method [25]. The Coulomb–Mohr hypothesis was adopted with a linear plastic flow law, as is recommended for the stability analyses of slopes, e.g., [26]. The vertical stress in the model increased linearly with depth. The Factor of Safety (FoS) was determined using the shear reduction method (SRM). The SRM is based on the assumption of an iterative reduction in the subsequent calculation steps of the strength parameters of the soil: friction angle and cohesion until the numerical stability of the model is achieved. The value of the FoS was calculated for the natural (dry model) and fully saturated (wet model) numerical models. The boundary conditions were introduced in such a way as to block the possibility of horizontal displacements on both side frames and vertical displacements on the bottom frame. The terrain surface in the numerical model was developed based on LIDAR data. The geometry of the geological structure was developed based on the extrapolation of data from the four geological profiles.

Analysis of the stability of the railway embankment slope was undertaken on the basis of geological-engineering documentation [27].

3. Results of Remote Sensing Test to Model the Morphology of the Area before and after the Landslide Occurred

To model the morphology before the landslide occurred, we used data available at www.geoportal.gov.pl (Figure 5). In this way, the morphology of the area before the landslide was reconstructed. As can be seen, the area is suitable for the accumulation of rainwater from trench No. 1. In location No. 2, the accumulation of the flowing water from trench No. 1 occurred, and a water-flow channel was created (Figure 6).

The digital terrain model after the landslide was developed using a terrestrial laser scanner is shown in Figure 6. Using the Riskan Pro Cloud Punk, the point clouds were combined in one area into a cloud of 3D points. Even at this analysis stage, a surface trough (No. 1) can be seen at the location of the landslide (Figure 7).

Figure 5. Digital model of the area obtained from LIDAR air photos (based on www.geoportal.gov.pl). Description: 1–trench used to collect rainwater; 2–the pass where rainwater flows out from the trench 1; 3–position of the railway track.

Figure 5. Digital model of the area obtained from LIDAR air photos (based on www.geoportal.gov.pl). Description: 1–trench used to collect rainwater; 2–the pass where rainwater flows out from the trench 1; 3–position of the railway track.

Figure 6. Narrow pass No. 2 resulted from the loss of the stability of the railway embankment (phot. Szwarkowski, D.).

Figure 6. Narrow pass No. 2 resulted from the loss of the stability of the railway embankment (phot. Szwarkowski, D.).

Figure 7. The digital model of the landslide consisting of cloud points obtained from terrestrial laser scanning.

Figure 7. The digital model of the landslide consisting of cloud points obtained from terrestrial laser scanning.

Based on the terrestrial measurements, a digital model of the ground surface was created (Figure 8). The obtained model allows the analysis of morphology changes in the context of the possible activation of the landslide. As a result of the water, the landslide main scarp formed, and a deeply grooved trough was created.

Figure 9 shows a model of the area after the landslide; the model was constructed using the data obtained from terrestrial laser scanning, which was superposed by the digital model obtained from the LIDAR data. The location of the object No. 2 from Figure 5 is probably where the landslide process was activated.

Figure 8. A digital model of the area based on terrestrial laser scanning.

Figure 8. A digital model of the area based on terrestrial laser scanning.

4. Results of Numerical Modelling

The parameters of each layer were obtained from geotechnical testing [27] (

Table 1. Parameters obtained from geotechnical testing [27].

Kind of Layer	Unit Weight γ [kN/m3]	Cohesion cu [kPa]	Internal Friction Angle φu [o]	Liquidity Index (IL) [-]	Filtration Coefficient k10 [m/s]
Crushed rock	16.0	1	40	-	10−3
Colluvium	18.5	5	29	-	10−4
Silty layer IIa	20.5	28	16	0.30	10−5
Silty layer IIb	21.0	32	19	0.18	10−7
Sandy silt layer III	21.5	32	20	0.10	10−7
Sandy layer IVa	19.5	1	32	-	10−3

).

Due to the assumed failure mechanism of the railway embankment, only characteristic parameters of the soil obtained from laboratory tests were used for the stability analysis. Because geotechnical studies were conducted after the landslide, it was not known precisely what the limits of the geotechnical layers were before the landslide (Figure 10). Cross-section II−II was the closest to the landslide main scarp. It was unknown whether the sandy layer was a continuous or a lens layer, as it would seem from geotechnical testing. The numerical model was based on the data obtained with some margin of uncertainty which was difficult to estimate. Figure 11 shows the geometry of numerical model of the area with the geotechnical layers. The load of the railway traffic was omitted in this numerical calculation, but it can be assumed to be a significant factor in activating the landslide process. The issue is important in the context of the impact of ground vibrations on the environment and anthropological constructions [28].

Two numerical models were analysed–dry and wet. The dry model was built on the assumed natural moisture of the layers, while the wet model represented fully saturated layers.

The results for calculating the stability for the dry model are shown in Figure 12. As can be seen, the obtained stability factor FoS = 1.17. Thus, it is close to the stability limit but was not exceeded. This explains why the slope did not lose stability until water influence changed the soil parameters.

Considering wet model assumptions, the IIA silty layer was plasticised. Liquidity index I_L = 0.75 was adopted. The strength parameter values were changed to φ_u = 11° and c_u = 6 kPa.

The results for calculating the stability for the wet model are shown in Figure 13. The obtained stability factor FoS = 0.9. This means the railway embankment lost its stability.

5. Discussion

LIDAR laser scanning allowed us to acquire data to model the area’s morphology before the landslide occurred. Analysing the morphology (Figure 5), a recess of the terrain is visible in the form of a trench (marked by 1). Rainwater from this trench flowed out through the trough marked by the number 2. Based on meteorological data, it can be assumed that precipitation reached about 100 mm per day within the few days before the occurrence of the landslide, and that the soil moisture due to the previous rainfall was several dozen centimetres. These data indicate–in the absence of water drainage on the right side of the embankment (Figure 5)–the possibility of a large waterway stagnation with a length of several dozen metres and a width of a dozen or so metres in the time immediately before the occurrence of the landslide. Such a volume of water would undoubtedly exert a large hydrostatic pressure on the ground−which had an increased pore pressure−and filtration for the aquatic layers of sand below. Analysing the system of geotechnical layers, silty layer IIa was the most likely to develop the slip surface (Figure 14). In addition, a sandy layer, IIb, located below layer IIa, constitutes a natural water flow path. The accumulation of excess water in the layer caused it to exert additional pressure on layer IIa. There was probably a hydrostatic piercing of waters from the aquatic layer IVa to the silty layer IIa. This caused the values of the strength parameters of silty layer IIa to deteriorate significantly due to the influence of the water. This follows from the fact that there is a small difference between the plasticity and liquidity limits in this type of soil in comparison with those of other cohesive soils [29]. With the increase in soil moisture, the attraction force between soil particles decreases and, therefore, the ground exhibits less coherence. In cohesive soils which have moisture close to the liquidity limit, the cohesion force disappears almost completely. As a consequence, the low cohesion of deeper layers may contribute to movements of larger masses of soils [30].

The formation of a slip surface was shown in a 3D numerical simulation. The dry model indicated the railway embankment was stable in soil parameters before the landslide occurred (FoS = 1.15). For the wet model, the embankment became unstable (FoS = 0.9). Additionally, stability coefficients were small prior to the landslide occurrence. According to Eurocode 7, the FoS coefficient should be at least 1.25 [31]. The calculated FoS did not comply with the safety condition stated in the standards; however, there was no failure because the FoS was greater than 1. The numerical calculation indicated that weakened silty layer IIa, an unfavourable morphology and the geological structure of the area were the main causes of the landslide. Based on direct observation of the area, the deterioration of the material parameters was caused by prolonged rainfall—results confirmed by the wet model calculation.

6. Conclusions

The study shows that the laser scanning data describe the scenario of the landslide process advantageously. Firstly, the model from the LIDAR data was used to reconstruct the morphology of the land prior to the landslide. This model indicated the unfavourable location of the area where water accumulated due to prolonged rainfall. Secondly, the model from terrestrial laser scanning showed the area morphology after the landslide and clearly illustrated the location of the landslide with the main scarp and the resulting water flows. As a result, it was shown that laser scanning might be applied to monitor transport route embankments and determine the causes of potential landslides.

The 3D numerical simulation confirmed the mechanism of the landslide process. The results significantly contributed to the identification of the causes underlying the landslide. The employed numerical models of the geological structure indicated the possibility of rainwater infiltration in the ground, which saturated the silty layer, thereby resulting in a rapid reduction in its shear strength.

To conclude, the study showed that the unfavourable morphology, geological structure of the soil mass and prolonged rainfall activated the landslide on the analysed railway track.